安装flash-Attention报错的解决编译flash-attn，python setup.py install

本文链接：https://blog.csdn.net/sinat_37574187/article/details/140684612

sudo docker run --gpus all -it --rm nvcr.io/nvidia/pytorch:24.06-py3

sudo docker run --gpus all -it --rm nvcr.io/nvidia/pytorch:24.06-py3

sudo docker run --gpus all -it --rm nvcr.io/nvidia/pytorch:24.06-py3
Unable to find image 'nvcr.io/nvidia/pytorch:24.06-py3' locally
24.06-py3: Pulling from nvidia/pytorch
a8b1c5f80c2d: Pulling fs layer
9c2a63d09670: Pulling fs layer
4f4fb700ef54: Pulling fs layer
c78c58a0000d: Waiting
ea63eb298c69: Waiting
aec371ddf6d1: Waiting
1925a16db8f6: Waiting
5767e2819d1e: Waiting
93fab9cd076a: Waiting
6bc5fc3bcef3: Waiting
a113fe9770cd: Waiting
aec371ddf6d1: Downloading [======>

编译flash-attn

[17/49] /usr/local/cuda/bin/nvcc --generate-dependencies-with-compile --dependency-output /data/zengzijian/Projects/InternVL-v2.0.0/flash-attention/build/temp.linux-x86_64-cpython-312/csrc/flash_attn/src/flash_fwd_split_hdim32_bf16_sm80.o.d -I/data/zengzijian/Projects/InternVL-v2.0.0/flash-attention/csrc/flash_attn -I/data/zengzijian/Projects/InternVL-v2.0.0/flash-attention/csrc/flash_attn/src -I/data/zengzijian/Projects/InternVL-v2.0.0/flash-attention/csrc/c